Is the buzzy deal a win for biopharma?
Earlier this month, AI player Databricks announced a takeover of database company Neon, in a bid to make itself more appealing to companies creating agentic bots. Databricks will integrate Neon's cloud-based platform, which is used by app and website developers including those in life sciences, to efficiently build and deploy agentic AI systems.
The deal, reportedly valued at about $1 billion, has generated its share of buzz in biopharma over the last two weeks, as do many tech pairings with the potential to accelerate the discovery-to-commercialization lifecycle.
“It would be absolutely great to be able to develop a lot more things more quickly and on the fly,” said Novartis’ Jen Yip, a 20-year life-sci veteran whose current role as clinical digital technology leader sees her focusing on harnessing tech to modernize drug development, including exploring practical AI use cases.
But there are operational hurdles to applying this technology in pharma, Yip cautioned, not to mention the industry’s traditional bias toward existing processes.
The potential
The AI in life sciences analytics market is projected to reach $3.6 billion globally by 2030, according to a report from The Research Insights. Databricks has already become a key player in that market by renting out analytics and other cloud-based apps that tap into AI-ready data to help enterprise tech systems run.
Neon, for its part, isn’t just used by human developers. AI bots also use its cloud-based database platform. Over 80% - or four out of every five - databases on Neon’s platform “are spun up by code, not humans,” noted Ali Ghodsi, Databricks co-founder and CEO, in a statement.
Databricks says the Neon deal addresses an oft-cited problem businesses face when getting AI agents off and running – syncing up the data these agents need in order to perform their services. By combining Neon’s serverless, relational database management system with its own services, Databricks says its customers will be able to deploy AI agents more efficiently.
To see how Neon’s database could underpin AI agents built on Databricks’ platform for life science clients, consider an area like clinical trial recruitment.
By deploying AI agents to reach out to identified patients within a very short time frame, they could speed up the contract identification and negotiation process. This could help keep leads for volunteers “warm,” by preventing those who respond to a pharma ad from falling through the cracks before signing up for the trial, said Yip.
These potential subjects “have a very pressing need,” she explained. “They are going to be looking everywhere [for a trial] and often take the first possibility that they get.”
By making better use of their clinical marketing, pharma companies could also relieve the burden on trial sites, which are notoriously overworked and understaffed (as is much of the medical profession). The fact that Neon databases can be created autonomously also could result in huge efficiencies.
The hurdles
That said, data limits could prevent drugmakers from realizing those agentic speed gains.
“Companies are hitting a lot of data-limitation walls already,” Yip said, “so a faster technology pulling on the same data may not necessarily speed things up there.”
Unless, of course, the AI agents were able to uncover novel data sources, which could expand the pool of potential patients.
But the biggest hurdle to overcome might be organizational resistance to breaking down some of the data silos that exist within pharma. Tech-driven acceleration is tough in an industry where many stakeholders prefer the status quo.
“A big challenge in big pharma is that organizationally - and in a lot of cases, mindset-wise - they are still set in a very old paradigm,” Yip observed. “There are still a lot of people who want to stay with the way things are already done.”
There’s internal politics - who's going to both fund the work and allow access? – as well as compliance, privacy and legal considerations. Onboarding external technology is notoriously hard.
Thus, while Databricks’ existing footprint in life science enhances its “trust-ability,” Yip said, the traditional hurdles can’t be ignored. Any pharma company would need to cultivate that trust.
Indeed, “Organizational capability as well as organizational trust are key to [setting up a] data foundation in AI,” said Yair Markovits, a partner at Beghou, which works with Databricks.
Building trust
Markovits, who made the remark during a March webinar which also involved fellow Beghou partner Amish Dhanani and Databricks solutions architect Baran Kavusturucu, MSc, counsels drugmakers to start with low-hanging fruit – i.e., low complexity/low impact tasks – before progressing along the continuum toward high-complexity/high-impact ones.
This continuum need not be a linear one. Organizations can progress in a non-linear way from one AI tool to the other, based on their internal approach to risk-taking, resources, timelines and trust, the speakers noted.
On the commercial analytics side, for instance, a company might deploy an LLM to tag insights in the notes of a medical science liaison (MSL) to make unstructured data into a structured format. The next level up would involve employing AI to summarize those insights. The highest-complexity/highest-impact task might entail building conversational AI agents that can answer questions about the insights accurately and thoroughly.
Of course, the AI model has to be safe and secure, with the right guardrails to ensure no personally identifiable information (PII) is at risk of being shared, and that the agents are doling out grounded responses versus hallucinations, added Kavusturucu.
Pharma knows it needs to work with AI. But just as organizational trust is key, lack thereof can lead to lower success.
“If there is no organizational trust and we have the best-in-class AI productivity tool, then nobody is going to use it,” added Markovits. “We need to build that trust in the data we acquire, in the technology we have, in the AI tools and how they were developed.”
The bottom line
AI agents hold a good deal of potential to tighten the relationship between clinical and commercial and, ultimately, the pathway to market. For this reason, there’s a lot of buzz around the Databricks + Neon deal.
Lest anyone think that solely marrying a relational database to agentic AI can supercharge drug development or marketing, though, there clearly are other considerations to take into account. Not the least of those in biopharma is the resistance to the adoption of technology – and the hefty trust deficit.
“There are a lot of scientist-types that will have to [be convinced] to let go of things,” said Yip. “It's a mindset thing that has to be considered in pharma.”
However, once organizational capability and organizational trust are in place, both of which are key to setting up the AI-ready data foundation, an initially reticent organization may often be more willing to move over to the strategic investment side, according to Markovits.
At the end of the day, the aforementioned data limits may prevent companies from fully realizing speed gains. But to the extent having Neon under its roof helps establish capability and trust, this deal could be a win for biopharma.
Can Databricks leverage Neon to build organizational capability and trust with biopharma so that the industry can more fully embrace agentic AI and speed up the clinical-to-commercial pathway? Are there other ways to overcome the trust deficit in onboarding new technology? Let me know your thoughts in the comments!
To watch the full Beghou/Databricks webinar click here.
As published here